Reservoir parameter prediction method and device

Abstract

The present invention provides a reservoir parameter prediction method and device, which relates to the field of geophysical technology. The method includes: obtaining a target function and training data of a neural network; the training data includes input training data and output training data; inputting the input training data into the target function, screening the first data output by the hidden layer of the target function, and obtaining the second data; generating the update weights between the hidden layer and the output layer of the target function according to the second data and the output training data; determining the updated target function according to the updated weights, and generating the reservoir parameter prediction results using the updated target function and the test data. The present invention can improve the computational efficiency of reservoir parameter prediction by updating the target function, and because the updated target function optimizes the weights in the target function, it can alleviate the overfitting of the neural network and improve the accuracy of the reservoir parameter prediction results.

Description

Technical Field

[0001] The present invention relates to the field of geophysical technology, and in particular to a reservoir parameter prediction method and device. Background Art

[0002] Reservoir characterization is an important part of the oil and gas exploration and development process. Various direct or indirect information (geological, seismic, logging data, etc.) signal processing and mathematical algorithms are used to obtain various underground formation parameters (elastic parameters, physical parameters), and the obtained formation parameters are used to describe the reservoir. Since seismic data can provide richer lateral information, seismic inversion technology has become one of the powerful means of reservoir description. However, due to the complexity of underground conditions and the uncertainty of seismic wave propagation underground, there is a complex nonlinear relationship between seismic data and formation parameters. Conventional seismic inversion technology usually uses mathematical methods to solve the formation parameters to describe the reservoir under certain assumptions.

[0003] Oil and gas reservoir characterization usually uses seismic data or logging data to invert the formation elastic parameters (including P-wave velocity, S-wave velocity, density, etc.), and then uses the obtained elastic parameters to invert the reservoir physical parameters that can characterize the reservoir characteristics (including lithofacies, porosity, shale content, water saturation, etc.). There is a nonlinear relationship between elastic parameters and physical parameters. It is difficult for traditional inversion methods based on theoretical models to accurately construct the nonlinear relationship between the two, resulting in large deviations in the inversion prediction results of physical parameters. Summary of the invention

[0004] The present invention provides a reservoir parameter prediction method and device, which can improve the computational efficiency of reservoir parameter prediction, alleviate the overfitting of neural networks, and improve the accuracy of prediction results.

[0005] In a first aspect, an embodiment of the present invention provides a reservoir parameter prediction method, which includes: obtaining an objective function and training data of a neural network; the training data includes input training data and output training data; inputting the input training data into the objective function, screening the first data output by the hidden layer of the objective function to obtain second data; generating update weights between the hidden layer and the output layer of the objective function according to the second data and the output training data; determining an updated objective function according to the updated weights, and generating a reservoir parameter prediction result using the updated objective function and test data.

[0006] In a second aspect, an embodiment of the present invention further provides a reservoir parameter prediction device, which includes: an acquisition module, used to acquire the objective function and training data of a neural network; the training data includes input training data and output training data; a processing module, used to input the input training data into the objective function, and filter the first data output by the hidden layer of the objective function to obtain second data; an update module, used to generate an update weight between the hidden layer and the output layer of the objective function according to the second data and the output training data; a result module, used to determine an updated objective function according to the updated weight, and generate a reservoir parameter prediction result using the updated objective function and test data.

[0007] In a third aspect, an embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned reservoir parameter prediction method when executing the computer program.

[0008] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program for executing the above-mentioned reservoir parameter prediction method.

[0009] The embodiments of the present invention bring the following beneficial effects: the embodiments of the present invention provide a reservoir parameter prediction scheme, which includes: obtaining the objective function and training data of a neural network; the training data includes input training data and output training data; inputting the input training data into the objective function, screening the first data output by the hidden layer of the objective function, and obtaining the second data; generating the update weights between the hidden layer and the output layer of the objective function according to the second data and the output training data; determining the updated objective function according to the updated weights, and generating the reservoir parameter prediction results using the updated objective function and the test data. The embodiments of the present invention can improve the computational efficiency of reservoir parameter prediction by updating the objective function, and because the updated objective function optimizes the weights in the objective function, the overfitting of the neural network can be alleviated, and the accuracy of the reservoir parameter prediction results can be improved.

[0010] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0011] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0013] Figure 1 A flow chart of a reservoir parameter prediction method provided by an embodiment of the present invention;

[0014] Figure 2 A flow chart of the reservoir parameter prediction method provided by the embodiment of the present invention;

[0015] Figure 3 The training data and prediction results provided by the embodiments of the present invention;

[0016] Figure 4 Test data and prediction results provided by the embodiments of the present invention;

[0017] Figure 5 A structural block diagram of a reservoir parameter prediction device provided by an embodiment of the present invention;

[0018] Figure 6 A structural block diagram of another reservoir parameter prediction device provided by an embodiment of the present invention;

[0019] Figure 7 A structural block diagram of a processing module provided in an embodiment of the present invention;

[0020] Figure 8 A structural block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] In recent years, machine learning methods have received extensive attention in the field of earth sciences and have achieved preliminary results in some fields. Existing methods use deep neural networks for seismic phase identification, and existing methods also use deep learning methods for reservoir parameter prediction. Existing methods also introduce deep learning methods into the velocity modeling process and achieve good results. The deep neural network algorithm is a machine learning algorithm. It uses optimization algorithms such as gradient descent to continuously optimize the parameters of the neural network and finally obtain a neural network that can characterize the relationship between input and output. This process will become more and more time-consuming as the number of layers and neurons in the network increases, and it is possible to fall into a local minimum during the solution process. The extreme learning machine method is a single-layer neural network algorithm. Compared with the deep neural network, it has a simpler network structure. In the process of network parameter optimization, the linear equation solution method is used instead of the iterative optimization method, and the optimization efficiency is improved. The extreme learning machine method belongs to a supervised learning algorithm. Like the deep neural network method, it has the phenomenon of overfitting. It has a good performance on the training set, but the effect on the test set is not good.

[0023] Based on this, a reservoir parameter prediction method and device provided in an embodiment of the present invention greatly improves the efficiency of reservoir parameter prediction in the reservoir characterization process, reduces manual workload, and improves the automation level and efficiency of reservoir characterization.

[0024] To facilitate understanding of this embodiment, a reservoir parameter prediction method disclosed in an embodiment of the present invention is first introduced in detail.

[0025] The present invention provides a reservoir parameter prediction method, see Figure 1 A flow chart of a reservoir parameter prediction method is shown, the method comprising the following steps:

[0026] S202, obtaining the objective function and training data of the neural network.

[0027] In the embodiment of the present invention, the objective function of the neural network is a pre-trained neural network prediction model. The training data includes input training data and output training data. The training data is used to further optimize the objective function.

[0028] It should be noted that the training data may be data obtained from seismic inversion or data obtained from well logging data in the target work area.

[0029] S204, inputting the input training data into the objective function, filtering the first data output by the hidden layer of the objective function, and obtaining second data.

[0030] In an embodiment of the present invention, the objective function includes an input layer, a hidden layer, and an output layer. The input training data is input into the objective function through the input layer, the weight parameters between the input layer and the hidden layer are calculated for the input training data, and the first data is output through the hidden layer. In order to reduce the influence of the overfitting of the neural network on the prediction results in the subsequent steps, part of the data is selected from the first data as the second data.

[0031] S206: Generate update weights between the hidden layer and the output layer of the objective function according to the second data and the output training data.

[0032] In the embodiment of the present invention, calculation is performed based on the second data and the output training data to optimize the weights between the hidden layer and the output layer of the objective function, thereby alleviating the overfitting of the neural network.

[0033] S208, determining an updated objective function according to the updated weights, and generating reservoir parameter prediction results using the updated objective function and the test data.

[0034] In the embodiment of the present invention, the original weights between the hidden layer and the output layer in the objective function are replaced by the updated weights to obtain an updated objective function. The test is input into the updated objective function to obtain the reservoir parameter prediction result.

[0035] It should be noted that if there is seismic data in the target work area, the prestack seismic inversion technology can be used to obtain the test data, or the test data can be determined based on the well logging data.

[0036] The embodiment of the present invention provides a reservoir parameter prediction scheme, which includes: obtaining the objective function and training data of a neural network; the training data includes input training data and output training data; inputting the input training data into the objective function, screening the first data output by the hidden layer of the objective function, and obtaining the second data; generating the update weights between the hidden layer and the output layer of the objective function according to the second data and the output training data; determining the updated objective function according to the updated weights, and generating the reservoir parameter prediction results using the updated objective function and the test data. The embodiment of the present invention can improve the computational efficiency of reservoir parameter prediction by updating the objective function, and because the updated objective function optimizes the weights in the objective function, it can alleviate the overfitting of the neural network and improve the accuracy of the reservoir parameter prediction results.

[0037] In order to improve the efficiency of reservoir parameter prediction, the following steps can be performed before obtaining the objective function of the neural network:

[0038] The objective function of the extreme learning machine neural network is constructed; the objective function is used to generate reservoir physical property parameters according to the formation elastic parameters.

[0039] In an embodiment of the present invention, an objective function that can be used to generate reservoir physical property parameters using formation elastic parameters is pre-constructed. The extreme learning machine method is a single-layer neural network algorithm. Compared with a deep neural network, it has a simpler network structure. In the process of network parameter optimization, a linear equation solving method is used instead of an iterative optimization method, which improves the optimization efficiency and thus improves the efficiency of reservoir parameter prediction.

[0040] In one embodiment, the formation elastic parameters include at least one or more of the P-wave velocity parameters, S-wave velocity parameters and density parameters; the reservoir physical property parameters include at least one or more of the lithofacies parameters, porosity parameters, mud content parameters and water saturation parameters.

[0041] In the embodiment of the present invention, an objective function is constructed to simultaneously invert discrete attributes (lithofacies) and continuous attributes (porosity, shale content, water saturation, etc.). As a type of supervised learning, the extreme learning machine method requires training data, including input data and corresponding output data. The input used in the embodiment of the present invention is P-wave velocity, S-wave velocity and density, and the output is lithofacies, porosity, shale content and water saturation.

[0042] In order to improve the accuracy of reservoir parameter prediction results and enhance the prediction ability of the model, the hidden layer includes multiple neurons; the first data output by the hidden layer of the objective function is screened to obtain the second data, which can be performed according to the following steps:

[0043] The first data is screened according to the contribution of multiple neurons to obtain second data.

[0044] In an embodiment of the present invention, considering that the contribution of each neuron in the hidden layer of the network to the output is different, if the probability of being discarded of each neuron is the same, this will cause the neurons that have a more important impact on the output result to be discarded during the training process, while the neurons that have a smaller impact on the output result are retained, which will affect the prediction ability of the algorithm to a certain extent. In this embodiment of the present invention, the first data is screened according to the contribution of the neurons to improve the accuracy of the reservoir parameter prediction results.

[0045] In order to improve the generalization ability of the method and the stability of the solution, the embodiment of the present invention adopts a biased discarding method to realize the regularization of the algorithm. Therefore, the first data includes the output values ​​of multiple neurons; the first data is screened according to the contribution of multiple neurons to obtain the second data, which can be performed according to the following steps:

[0046] A threshold is determined according to a plurality of output values; mask data is determined according to the output values ​​and the threshold; and the mask data is used to filter the first data to obtain the second data.

[0047] In the embodiment of the present invention, the output values ​​of each neuron in the hidden layer are first counted, and multiple output values ​​of multiple neurons are calculated to obtain a threshold value for determining the probability of neuron retention. After obtaining the threshold value, the probability information of neuron retention is determined based on the output value and the threshold value, and then the mask data is determined based on the probability information. In the embodiment of the present invention, the mask data can be binary mask data.

[0048] It should be noted that the mask can use the selected image, graphic or object to block the processed image (all or part) to control the area or process of image processing. Specifically, it can be used to extract the region of interest: multiply the pre-made region of interest mask with the image to be processed to obtain the region of interest image, the image value in the region of interest remains unchanged, and the image value outside the region is 0; shielding effect: use the mask to shield certain areas on the image so that it does not participate in the processing or the calculation of the processing parameters, or only the shielded area is processed or counted; structural feature extraction: use similarity variables or image matching methods to detect and extract structural features in the image that are similar to the mask; production of special shape images.

[0049] In order to make a more accurate judgment on the contribution of neurons, the threshold can be determined based on multiple output values. The following steps can be followed:

[0050] Calculate the average of multiple output values ​​and use the average as the threshold.

[0051] In the embodiment of the present invention, the average value of multiple output values ​​is used as the threshold value. In the actual implementation process, other calculation methods can also be used to process the multiple output values. The calculation can be performed according to actual needs, and the embodiment of the present invention does not make specific limitations on this.

[0052] In one embodiment, the first data is screened using the mask data according to the following formula to obtain the second data: do =A(WX)·M, where A do is the second data, · is the Hadamard product, M is the mask, W is the weight between the input layer and the hidden layer of the objective function, X is the input training data of the objective function, and A is the first data.

[0053] In one embodiment, the update weights between the hidden layer and the output layer of the objective function are generated according to the second data and the output training data according to the following formula: Among them, Θ * To update the weight, A do is the second data, and Y is the output training data.

[0054] See also Figure 2 The implementation flow chart of the reservoir parameter prediction method is shown in the figure. The implementation process of the method is described below with a specific embodiment.

[0055] Figure 2 The technical process of predicting reservoir parameters according to elastic parameters or logging data obtained by seismic inversion in an embodiment of the present invention is given. The method can be performed according to the following steps:

[0056] Step 101: First, the logging data in the target work area are divided into training data and test data, and the input data are determined to be elastic parameters, including P-wave velocity, S-wave velocity and density; the output data are determined to be reservoir physical parameters, including lithofacies, porosity, shale content and water saturation.

[0057] Step 102: First, according to the data, determine the number of hidden layer neurons as H, and generate random numbers to assign weights and bias items connecting the input and hidden layer neurons respectively; then, input the input data into the network to obtain the output of the hidden layer neurons, and count the output of the hidden layer neurons to obtain the mean as the discard threshold of the biased discard regularization method.

[0058] Step 103: Based on the network structure determined in the previous step and the threshold for applying the biased dropout method, the training data is used for complete training to obtain the final network model.

[0059] Step 104: If there is seismic data in the work area, the pre-stack seismic inversion technology can be used to obtain the P-wave velocity, S-wave velocity and density, and the inversion result is used as the input of the network; if the interpretation is for well data, the P-wave velocity, S-wave velocity and density in the logging data are used as the input of the network.

[0060] Step 105: The P-wave velocity, S-wave velocity and density obtained in step 104 are used as input data of the input network, and finally the lithofacies, porosity, shale content and water saturation are directly predicted.

[0061] The embodiment of the present invention first determines the extreme learning machine network model structure according to the characteristics of the input data, so as to achieve the direct and simultaneous prediction of lithofacies, porosity, shale content and water saturation, thereby improving the efficiency of reservoir parameter prediction; then, the biased dropout method is introduced as a regularization constraint in the network training process to improve the stability and generalization ability of the algorithm, thereby obtaining reliable reservoir parameter prediction results. The main contents of the embodiment of the present invention include the following two parts:

[0062] (1) Construct an objective function that simultaneously inverts discrete attributes (lithofacies) and continuous attributes (porosity, shale content, water saturation, etc.).

[0063] The input data is represented as

[0064] X={x n ,n=1,…,N} (1)

[0065] x n =[x 1n ,…,x Dn ] T (2)

[0066] Where N represents the number of training data samples, and D represents the number of attributes of each input sample. The parameters to be determined include lithofacies, porosity, shale content, and water saturation, which are represented by f, Φ, Vsh, and Sw, respectively. When there are C different lithofacies categories of lithofacies parameters to be determined, the present invention introduces a matrix F to represent the lithofacies of each data sample, and the size of the matrix is ​​C×N. n For example, if the lithofacies is type C, then F cn is 1; otherwise, the other elements are 0. In order to obtain the above four parameters at the same time, the four parameters are combined to construct a matrix Y of size (C+3)×N. When the number of hidden layer neurons used is H, the weight connecting the input data and the hidden layer neurons is The bias term is The weights connecting the hidden neurons and the output are Where W and b are random values. By solving formula (3), we can get Θ.

[0067] θ=(AA T ) -1 AY T (3)

[0068] in,

[0069]

[0070] The matrix A is the output of the hidden layer, a(·) is the selected activation function, and the present invention adopts the ReLU activation function, w h represents the hth row vector of W. Then the predicted value Y * It can be solved using the following formula.

[0071] Y * =Θ T A (5)

[0072] (2) The embodiment of the present invention selects the mean value of each neuron output value as the threshold. The probability of a neuron output value greater than the threshold being discarded is relatively small, and vice versa, as shown in the following formula:

[0073]

[0074] Among them, p is the probability of a neuron being retained, p1>p2, and a is the threshold. The weight θ between the hidden layer neurons and the output is solved using the training data * It can be expressed as

[0075]

[0076] A do =A(WX)·M (8)

[0077] Among them, A do is the neuron output matrix obtained after applying the dropout method, · ​​is the Hadamard product, M is a binary mask, one of its elements has a probability of p being assigned to 1, and (1-p) is assigned to 0, and p follows the Bernoulli distribution. Then, using the weight Θ obtained by solving * And matrix A to obtain the predicted values ​​(lithology, porosity, shale content and water saturation).

[0078] See also Figure 3 The training data and prediction results shown are Figure 3 (a) Figure 3 (b) Figure 3 (c) in the equation represents the input longitudinal wave velocity, shear wave velocity and density, respectively. Figure 3 (d) Figure 3 (e) in the figure represents the actual value and predicted value of porosity, Figure 3 (f) Figure 3 The (g) in the equation represents the actual value and predicted value of the mud content, Figure 3 (h) Figure 3 The (i) in the equation represents the actual value and predicted value of water saturation, Figure 3 (j) Figure 3 (k) Figure 3 The (l) in the figure represents the error between the predicted value and the true value of porosity, shale content and water saturation. The error is small. Figure 3 (m) Figure 3 The (n) in the figure represent the true value and predicted value of the lithofacies, respectively.

[0079] See also Figure 4 The test data and prediction results shown are Figure 4 (a) Figure 4 (b) Figure 4 (c) in the equation represents the input longitudinal wave velocity, shear wave velocity and density, respectively. Figure 4 (d) Figure 4 (e) in the figure represents the actual value and predicted value of porosity, Figure 4 (f) Figure 4 The (g) in the equation represents the actual value and predicted value of the mud content, Figure 4 (h) Figure 4 The (i) in the equation represents the actual value and predicted value of water saturation, Figure 4 (j) Figure 4 (k) Figure 4 The (l) in the figure represents the error between the predicted value and the true value of porosity, shale content and water saturation. The error is small. Figure 4 (m) Figure 4 The (n) in the figure represents the true value and predicted value of the lithofacies. The test data experiment results show that the method proposed in the embodiment of the present invention can achieve a better prediction effect.

[0080] The present invention provides a reservoir parameter prediction method and device. The method can simultaneously predict lithofacies, porosity, mud content and water saturation by using elastic parameters obtained from seismic data or well logging data, thereby greatly improving efficiency. A biased discarding method is used to realize regularization of network training, thereby improving the generalization ability of the trained model. The method only requires one training, thereby avoiding the problem of excessive time-consuming iterative optimization process during neural network training, thereby improving parameter prediction efficiency.

[0081] The present invention also provides a reservoir parameter prediction device, as described in the following embodiments. Since the principle of the device to solve the problem is similar to that of the reservoir parameter prediction method, the implementation of the device can refer to the implementation of the reservoir parameter prediction method, and the repeated parts will not be repeated. Figure 5 A structural block diagram of a reservoir parameter prediction device is shown, the device comprising:

[0082] An acquisition module 71 is used to acquire the objective function and training data of the neural network; the training data includes input training data and output training data; a processing module 72 is used to input the input training data into the objective function, filter the first data output by the hidden layer of the objective function, and obtain the second data; an update module 73 is used to generate the update weights between the hidden layer and the output layer of the objective function according to the second data and the output training data; a result module 74 is used to determine the updated objective function according to the updated weights, and generate reservoir parameter prediction results using the updated objective function and the test data.

[0083] In one embodiment, see Figure 6 Another structural block diagram of a reservoir parameter prediction device is shown, the device also includes a construction module 75, which is used to: construct an objective function of an extreme learning machine neural network; the objective function is used to generate reservoir physical property parameters according to formation elastic parameters.

[0084] In one embodiment, the formation elastic parameters include at least one or more of the P-wave velocity parameters, S-wave velocity parameters and density parameters; the reservoir physical property parameters include at least one or more of the lithofacies parameters, porosity parameters, mud content parameters and water saturation parameters.

[0085] In one embodiment, the hidden layer includes a plurality of neurons; and the processing module is specifically used to: filter the first data according to the contribution of the plurality of neurons to obtain the second data.

[0086] In one embodiment, the first data includes output values ​​of a plurality of neurons; see Figure 7 As shown in the structural block diagram of the processing module, the processing module includes: a threshold unit 721, used to determine a threshold according to multiple output values; a data unit 722, used to determine mask data according to the output value and the threshold; and a screening unit 723, used to screen the first data using the mask data to obtain the second data.

[0087] In one embodiment, the threshold unit is specifically used to calculate an average value of multiple output values ​​and use the average value as the threshold value.

[0088] In one embodiment, the data unit is specifically used to: use the mask data to filter the first data according to the following formula to obtain the second data: A do =A(WX)·M, where A do is the second data, · is the Hadamard product, M is the mask, W is the weight between the input layer and the hidden layer of the objective function, X is the input training data of the objective function, and A is the first data.

[0089] In one embodiment, the screening unit is specifically used to generate an update weight between a hidden layer and an output layer of the objective function according to the second data and the output training data according to the following formula: Among them, Θ * To update the weight, A do is the second data, and Y is the output training data.

[0090] The embodiment of the present invention also provides a computer device, see Figure 8 The computer device shown in the schematic block diagram includes a memory 81, a processor 82, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned reservoir parameter prediction methods are implemented.

[0091] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the computer device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0092] An embodiment of the present invention further provides a computer-readable storage medium, which stores a computer program for executing any of the above-mentioned reservoir parameter prediction methods.

[0093] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0094] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0095] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0096] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0097] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A reservoir parameter prediction method, characterized in that: include: Obtain the objective function and training data of the neural network; The training data includes input training data and output training data; Inputting the input training data into the objective function, filtering the first data output by the hidden layer of the objective function to obtain second data; generating an update weight between the hidden layer and the output layer of the objective function according to the second data and the output training data; Determining an updated objective function according to the updated weights, and generating reservoir parameter prediction results using the updated objective function and test data; The hidden layer includes a plurality of neurons; Screening the first data output by the hidden layer of the objective function to obtain the second data includes: Filtering the first data according to the contribution of the multiple neurons to obtain second data; The first data includes output values ​​of the plurality of neurons; The first data is screened according to the contribution of the plurality of neurons to obtain second data, including: determining a threshold value according to a plurality of said output values; determining mask data according to the output value and the threshold; The first data is screened using the mask data to obtain second data.

2. The method according to claim 1, characterized in that Before obtaining the objective function of the neural network, it also includes: An objective function of an extreme learning machine neural network is constructed; the objective function is used to generate reservoir physical property parameters according to formation elastic parameters.

3. The method according to claim 2, characterized in that The formation elastic parameters include at least one or more of the following: P-wave velocity parameters, S-wave velocity parameters and density parameters; the reservoir physical property parameters include at least one or more of the following: lithofacies parameters, porosity parameters, mud content parameters and water saturation parameters.

4. The method according to claim 1, characterized in that: Determining a threshold value according to a plurality of the output values ​​comprises: An average value of the plurality of output values ​​is calculated, and the average value is used as a threshold value.

5. The method according to claim 1, characterized in that include: The first data is screened using the mask data according to the following formula to obtain the second data: A do =A(WX)·M Among them, A do is the second data, · is the Hadamard product, M is the mask, W is the weight between the input layer and the hidden layer of the objective function, X is the input training data of the objective function, and A is the first data.

6. The method according to claim 1, characterized in that include: The update weight between the hidden layer and the output layer of the objective function is generated according to the second data and the output training data according to the following formula: Among them, Θ * To update the weight, A do is the second data, and Y is the output training data.

7. A reservoir parameter prediction device, characterized in that: include: An acquisition module is used to obtain the objective function and training data of the neural network; The training data includes input training data and output training data; A processing module, configured to input the input training data into the objective function, filter the first data output by the hidden layer of the objective function, and obtain second data; An updating module, configured to generate an updated weight between the hidden layer and the output layer of the objective function according to the second data and the output training data; A result module, used for determining an updated objective function according to the updated weights, and generating a reservoir parameter prediction result using the updated objective function and test data; The hidden layer includes a plurality of neurons; the processing module is specifically used for: Filtering the first data according to the contribution of the multiple neurons to obtain second data; The first data includes output values ​​of the plurality of neurons; The processing module comprises: A threshold unit, used for determining a threshold according to a plurality of said output values; A data unit, used for determining mask data according to the output value and the threshold value; A screening unit is used to screen the first data using the mask data to obtain second data.

8. The device according to claim 7, characterized in that Also included are building blocks for: An objective function of an extreme learning machine neural network is constructed; the objective function is used to generate reservoir physical property parameters according to formation elastic parameters.

9. The device according to claim 8, characterized in that The formation elastic parameters include at least one or more of the following: P-wave velocity parameters, S-wave velocity parameters and density parameters; the reservoir physical property parameters include at least one or more of the following: lithofacies parameters, porosity parameters, mud content parameters and water saturation parameters.

10. The device according to claim 7, characterized in that The threshold unit is specifically used for: An average value of the plurality of output values ​​is calculated, and the average value is used as a threshold value.

11. The device according to claim 7, characterized in that The data unit is specifically used for: The first data is screened using the mask data according to the following formula to obtain the second data: A do =A(WX)·M Among them, A do is the second data, · is the Hadamard product, M is the mask, W is the weight between the input layer and the hidden layer of the objective function, X is the input training data of the objective function, and A is the first data.

12. The device according to claim 7, characterized in that The screening unit is specifically used for: The update weight between the hidden layer and the output layer of the objective function is generated according to the second data and the output training data according to the following formula: Among them, Θ * To update the weight, A do is the second data, and Y is the output training data.

13. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the reservoir parameter prediction method according to any one of claims 1 to 6 is implemented.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the reservoir parameter prediction method according to any one of claims 1 to 6 is implemented.

Images